[WM] Webmake performance

Justin Mason jm.jmason.org
Thu Nov 21 10:49:03 GMT 2002


Oliver Strutynski said:
> BTW: Is it really a good idea to compare the generated file with the
> existin g output file? It probably is pretty time-consuming and since
> webmake usually does a good job with dependency analysis it should only
> process a file it it has changed. That leaves the case where a file has
> changed (or has a new time-stamp) but the output remains the same. Is
> this really a common case?

Well, the idea is to save bandwidth as well -- by not changing the file,
the file's modification date remains unchanged, so web caches and browsers
using If-Modified-Since headers etc. will not invalidate it.

However you may have a point regarding how much time it takes to do this.
I've been considering a good feature might be a "quick rebuild" flag
which turns this off, and also regenerates the files in place (no ".new"
suffixes which are renamed quickly at the end) so that files can be
viewed before the webmake command has finished running completely.
But time to implement it is in short supply :(

> Now for getting file.section, the file is opened and parsed. The dbg
> message is "subst from eval code (always rebuilt)". This should not
> happen as the system knows that the files have no external dependencies.
> Later when lookin g at the files in all_files in a FOR-loop  webmake
> correctly sees that the files have not been changed and have no external
> dependencies.
> 
> Any way to work around that?

no, afraid not... this code is a bit crufty these days.  Again this needs
a solid rewrite some day soon...

as a matter of interest, is the "section" metadata stored in the files
using <wmmeta>, or external to them using <metatable>?

--j.




More information about the Webmake-talk mailing list